Hugging Face's logo

qualcomm
/
Midas-V2-Quantized

Depth Estimation
PyTorch
LiteRT
ONNX
quantized
android
Model card Files Community

Midas-V2-Quantized: Optimized for Mobile Deployment

Quantized Deep Convolutional Neural Network model for depth estimation

Midas is designed for estimating depth at each point in an image.

This model is an implementation of Midas-V2-Quantized found here.

This repository provides scripts to run Midas-V2-Quantized on Qualcomm® devices. More details on model performance across various devices, can be found here.

Model Details

  • Model Type: Depth estimation
  • Model Stats:
    • Model checkpoint: MiDaS_small
    • Input resolution: 256x256
    • Number of parameters: 16.6M
    • Model size: 16.6 MB
Model Device Chipset Target Runtime Inference Time (ms) Peak Memory Range (MB) Precision Primary Compute Unit Target Model
Midas-V2 Samsung Galaxy S23 Snapdragon® 8 Gen 2 TFLITE 1.061 ms 0 - 133 MB INT8 NPU Midas-V2-Quantized.tflite
Midas-V2 Samsung Galaxy S23 Snapdragon® 8 Gen 2 QNN 1.344 ms 0 - 134 MB INT8 NPU Midas-V2-Quantized.so
Midas-V2 Samsung Galaxy S23 Snapdragon® 8 Gen 2 ONNX 114.034 ms 0 - 138 MB INT8 NPU Midas-V2-Quantized.onnx
Midas-V2 Samsung Galaxy S24 Snapdragon® 8 Gen 3 TFLITE 0.753 ms 0 - 55 MB INT8 NPU Midas-V2-Quantized.tflite
Midas-V2 Samsung Galaxy S24 Snapdragon® 8 Gen 3 QNN 0.969 ms 0 - 52 MB INT8 NPU Midas-V2-Quantized.so
Midas-V2 Samsung Galaxy S24 Snapdragon® 8 Gen 3 ONNX 96.517 ms 21 - 353 MB INT8 NPU Midas-V2-Quantized.onnx
Midas-V2 Snapdragon 8 Elite QRD Snapdragon® 8 Elite TFLITE 0.598 ms 0 - 32 MB INT8 NPU Midas-V2-Quantized.tflite
Midas-V2 Snapdragon 8 Elite QRD Snapdragon® 8 Elite QNN 0.88 ms 0 - 30 MB INT8 NPU Use Export Script
Midas-V2 Snapdragon 8 Elite QRD Snapdragon® 8 Elite ONNX 82.409 ms 13 - 334 MB INT8 NPU Midas-V2-Quantized.onnx
Midas-V2 SA7255P ADP SA7255P TFLITE 10.329 ms 0 - 27 MB INT8 NPU Midas-V2-Quantized.tflite
Midas-V2 SA7255P ADP SA7255P QNN 11.549 ms 0 - 10 MB INT8 NPU Use Export Script
Midas-V2 SA8255 (Proxy) SA8255P Proxy TFLITE 1.068 ms 0 - 133 MB INT8 NPU Midas-V2-Quantized.tflite
Midas-V2 SA8255 (Proxy) SA8255P Proxy QNN 1.285 ms 0 - 2 MB INT8 NPU Use Export Script
Midas-V2 SA8295P ADP SA8295P TFLITE 1.918 ms 0 - 30 MB INT8 NPU Midas-V2-Quantized.tflite
Midas-V2 SA8295P ADP SA8295P QNN 2.223 ms 0 - 18 MB INT8 NPU Use Export Script
Midas-V2 SA8650 (Proxy) SA8650P Proxy TFLITE 1.069 ms 0 - 133 MB INT8 NPU Midas-V2-Quantized.tflite
Midas-V2 SA8650 (Proxy) SA8650P Proxy QNN 1.286 ms 0 - 4 MB INT8 NPU Use Export Script
Midas-V2 SA8775P ADP SA8775P TFLITE 1.57 ms 0 - 27 MB INT8 NPU Midas-V2-Quantized.tflite
Midas-V2 SA8775P ADP SA8775P QNN 1.8 ms 0 - 10 MB INT8 NPU Use Export Script
Midas-V2 RB3 Gen 2 (Proxy) QCS6490 Proxy TFLITE 3.681 ms 0 - 44 MB INT8 NPU Midas-V2-Quantized.tflite
Midas-V2 RB3 Gen 2 (Proxy) QCS6490 Proxy QNN 5.84 ms 0 - 11 MB INT8 NPU Use Export Script
Midas-V2 RB5 (Proxy) QCS8250 Proxy TFLITE 15.724 ms 0 - 2 MB INT8 NPU Midas-V2-Quantized.tflite
Midas-V2 QCS8275 (Proxy) QCS8275 Proxy TFLITE 10.329 ms 0 - 27 MB INT8 NPU Midas-V2-Quantized.tflite
Midas-V2 QCS8275 (Proxy) QCS8275 Proxy QNN 11.549 ms 0 - 10 MB INT8 NPU Use Export Script
Midas-V2 QCS8550 (Proxy) QCS8550 Proxy TFLITE 1.067 ms 0 - 132 MB INT8 NPU Midas-V2-Quantized.tflite
Midas-V2 QCS8550 (Proxy) QCS8550 Proxy QNN 1.286 ms 0 - 3 MB INT8 NPU Use Export Script
Midas-V2 QCS9075 (Proxy) QCS9075 Proxy TFLITE 1.57 ms 0 - 27 MB INT8 NPU Midas-V2-Quantized.tflite
Midas-V2 QCS9075 (Proxy) QCS9075 Proxy QNN 1.8 ms 0 - 10 MB INT8 NPU Use Export Script
Midas-V2 QCS8450 (Proxy) QCS8450 Proxy TFLITE 1.548 ms 0 - 42 MB INT8 NPU Midas-V2-Quantized.tflite
Midas-V2 QCS8450 (Proxy) QCS8450 Proxy QNN 1.815 ms 0 - 47 MB INT8 NPU Use Export Script
Midas-V2 Snapdragon X Elite CRD Snapdragon® X Elite QNN 1.434 ms 0 - 0 MB INT8 NPU Use Export Script
Midas-V2 Snapdragon X Elite CRD Snapdragon® X Elite ONNX 108.214 ms 66 - 66 MB INT8 NPU Midas-V2-Quantized.onnx

Installation

Install the package via pip:

pip install "qai-hub-models[midas-quantized]"

Configure Qualcomm® AI Hub to run this model on a cloud-hosted device

Sign-in to Qualcomm® AI Hub with your Qualcomm® ID. Once signed in navigate to Account -> Settings -> API Token.

With this API token, you can configure your client to run models on the cloud hosted devices.

qai-hub configure --api_token API_TOKEN

Navigate to docs for more information.

Demo off target

The package contains a simple end-to-end demo that downloads pre-trained weights and runs this model on a sample input.

python -m qai_hub_models.models.midas_quantized.demo

The above demo runs a reference implementation of pre-processing, model inference, and post processing.

NOTE: If you want running in a Jupyter Notebook or Google Colab like environment, please add the following to your cell (instead of the above).

%run -m qai_hub_models.models.midas_quantized.demo

Run model on a cloud-hosted device

In addition to the demo, you can also run the model on a cloud-hosted Qualcomm® device. This script does the following:

  • Performance check on-device on a cloud-hosted device
  • Downloads compiled assets that can be deployed on-device for Android.
  • Accuracy check between PyTorch and on-device outputs.
python -m qai_hub_models.models.midas_quantized.export

Profiling Results
------------------------------------------------------------
Midas-V2
Device                          : Samsung Galaxy S23 (13)
Runtime                         : TFLITE                 
Estimated inference time (ms)   : 1.1                    
Estimated peak memory usage (MB): [0, 133]               
Total # Ops                     : 140                    
Compute Unit(s)                 : NPU (140 ops)

How does this work?

This export script leverages Qualcomm® AI Hub to optimize, validate, and deploy this model on-device. Lets go through each step below in detail:

Step 1: Compile model for on-device deployment

To compile a PyTorch model for on-device deployment, we first trace the model in memory using the jit.trace and then call the submit_compile_job API.

import torch

import qai_hub as hub
from qai_hub_models.models.midas_quantized import Model

# Load the model
torch_model = Model.from_pretrained()

# Device
device = hub.Device("Samsung Galaxy S24")

# Trace model
input_shape = torch_model.get_input_spec()
sample_inputs = torch_model.sample_inputs()

pt_model = torch.jit.trace(torch_model, [torch.tensor(data[0]) for _, data in sample_inputs.items()])

# Compile model on a specific device
compile_job = hub.submit_compile_job(
    model=pt_model,
    device=device,
    input_specs=torch_model.get_input_spec(),
)

# Get target model to run on-device
target_model = compile_job.get_target_model()

Step 2: Performance profiling on cloud-hosted device

After compiling models from step 1. Models can be profiled model on-device using the target_model. Note that this scripts runs the model on a device automatically provisioned in the cloud. Once the job is submitted, you can navigate to a provided job URL to view a variety of on-device performance metrics.

profile_job = hub.submit_profile_job(
    model=target_model,
    device=device,
)

Step 3: Verify on-device accuracy

To verify the accuracy of the model on-device, you can run on-device inference on sample input data on the same cloud hosted device.

input_data = torch_model.sample_inputs()
inference_job = hub.submit_inference_job(
    model=target_model,
    device=device,
    inputs=input_data,
)
    on_device_output = inference_job.download_output_data()

With the output of the model, you can compute like PSNR, relative errors or spot check the output with expected output.

Note: This on-device profiling and inference requires access to Qualcomm® AI Hub. Sign up for access.

Deploying compiled model to Android

The models can be deployed using multiple runtimes:

  • TensorFlow Lite (.tflite export): This tutorial provides a guide to deploy the .tflite model in an Android application.

  • QNN (.so export ): This sample app provides instructions on how to use the .so shared library in an Android application.

View on Qualcomm® AI Hub

Get more details on Midas-V2-Quantized's performance across various devices here. Explore all available models on Qualcomm® AI Hub

License

  • The license for the original implementation of Midas-V2-Quantized can be found here.
  • The license for the compiled assets for on-device deployment can be found here

References

Community

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
Depth Estimation
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support